3D-Printed Synthetic Vocal Fold Models

Abstract

Objective:

Synthetic vocal fold (VF) models used for studying the physics of voice production are comprised of silicone and fabricated using traditional casting processes. The purpose of this study was to develop and demonstrate a new method of creating synthetic VF models through 3D printing in order to reduce model fabrication time, increase yield, and lay the foundation for future models with more lifelike geometric, material, and vibratory properties.

Study Design:

Basic science

Methods:

A 3D printing technique based on embedding a UV-curable liquid silicone into a gel-like medium was selected and refined. Cubes were printed and subjected to tensile testing to characterize their material properties. Self-oscillating vocal fold models were then printed, coated with a thin layer of silicone representing the epithelium, and used in phonation tests to gather onset pressure, frequency, and amplitude data.

Results:

The cubes were found to be anisotropic, exhibiting different modulus values depending on the orientation of the printed layers. The VF models self-oscillated and withstood the strains induced by phonation. Print parameters were found to affect model vibration frequency and onset pressure. Primarily due to the design of the VF models, their onset pressures were higher than what is found in human VFs. However, their frequencies were within a comparable range.

Conclusion:

The results demonstrate the ability to 3D print synthetic, self-oscillating VF models. It is anticipated that this method will be further refined and used in future studies exploring flow-induced vibratory characteristics of phonation.

INTRODUCTION

Synthetic vocal fold (VF) models play an important role in studying voice production. These models are typically made of silicone and are intended to mimic the geometry and material properties of human VFs. Relative to in vivo and excised larynx experiments, synthetic model experiments are more readily accessible. These models allow material properties and model geometries to be parametrically varied in order to observe the effects of these changes on the models’ flow-induced vibratory responses. They have been used in various applications, such as estimating VF collision forces [1], studying the influence of VF asymmetry on phonation [2–4], and developing contact probes [5] (see [6] for a review of synthetic VF models).

Synthetic VF models are typically fabricated through a traditional casting process that involves multiple steps. A geometric computer model is first created, from which a positive mold is fabricated and used to create a negative mold. The VF model is then cast in the mold using liquid silicone. After curing, the model is then carefully removed from the mold. Some models [7] consist of multiple layers of silicone of differing stiffness to imitate the layered structure of human VF tissue, in which case this casting process is repeated using additional molds.

This process has a few inherent limitations. The process itself is time consuming, and creating a model with new geometry is time intensive. Models often tear or otherwise fail when being removed from their molds. Consequently, significant time is required to create these models. Further, the nature of casting imposes constraints on geometries that can be fabricated. Yet another disadvantage is the necessity for release agent to be used to facilitate removal of cured silicone molds. This release agent can inhibit curing of the silicone for very soft layers, and further, unpublished studies have suggested that it may be cytotoxic, thereby potentially limiting the use of some VF models in tissue engineering applications.

Notwithstanding these limitations, researchers have continued to use VF models because of their advantages. However, improved fabrication processes, in addition to the development of models with more lifelike characteristics, remain desirable. The purpose of this research was to apply ultra-soft 3D printing techniques to improve the fabrication of synthetic silicone VF models and enhance their potential for a wider range of applications. This was accomplished by building a 3D printer capable of extruding liquid silicone into a support matrix, identifying a silicone material that was both 3D printable and had mechanical properties suitable for voice research, tuning the printer and control software to achieve the best overall prints, and validating the printer, silicone, and printing process by fabricating and testing 3D-printed models.

In the following sections, the 3D printer and selected silicone are summarized. Material properties measured using 3D-printed cubes are given and compared to the properties of human VF tissue. Printed, self-oscillating VF models are introduced, and the results of tests to measure onset pressure, vibratory frequency, and model motion are reported. Finally, capabilities, limitations, and areas for future research are discussed. As a whole, the following sections contain an overview of this new approach for fabricating VF models, and for additional details and further study, the reader is referred to Romero [8].

METHODS

Silicone Printing Method

3D Printing Process

A printing method was selected in which 3D-printed structures are formed by injecting liquid silicone into a support matrix. Variations of this general approach appear in the literature, e.g. [9–12]. The particular method selected here was based on that described by O’Bryan et al. [13,14], in which a blunt-tip needle attached to a syringe that has been filled with uncured liquid silicone is inserted into the support matrix. The silicone is then extruded through the translating needle into the support matrix layer-by-layer as illustrated in Fig. 1. The support matrix holds the liquid silicone in place, maintaining its desired shape until the silicone is cured.

FIGURE 1.

Diagram of 3D printing process used in the present study. The needle traverses through the support matrix, extruding liquid silicone layer-by-layer. The support matrix keeps the deposited liquid silicone in place until it can be subsequently cured and removed from the support matrix.

Support Matrix

The formulation of the support matrix is dependent on the material being embedded, although many support matrices have similar rheological properties. Most matrices use microparticles to create a gel-like slurry that behaves as a solid at low shear stress and as a fluid at high stress [9]. In the presently-used formulation, based on [13], the support matrix is prepared by mixing 95.5 wt% mineral oil with two copolymers (G1702 and G1650, Kraton Corporation) at 2.25 wt% each and then stirred at 100°C for four to six hours [13]. Once the polymers have mixed entirely, the matrix is placed into individual printing containers while still warm, degassed to remove air bubbles, and allowed to cool to room temperature for printing.

Silicone

The UV-curable silicone elastomer Silopren UV Electro 225 (hereafter referred to as UV Electro; Momentive Performance Materials, Inc., Waterford, NY) was used as the extruded silicone. UV Electro consists of a silicone base combined with a UV curing agent. The base and UV curing agent were mixed at a 10:1 ratio by weight. In the present studies, UV Electro cured in approximately 10 minutes in a countertop UV curing bed (Dreve Polylux 2000), which is significantly faster than Ecoflex 00–30 (Smooth-On, Inc., Macungie, PA), a two-part RTV silicone that has been used for VF model fabrication in numerous studies [2,4,5,7,15–18]. Further, two-part RTV silicones such as Ecoflex begin curing once they are combined, which means their usable time while uncured is limited. The use of a UV-curable silicone eliminates the possibility of silicone curing in the syringe during printing.

As with Ecoflex, silicone thinner can be added to UV Electro to reduce its cured stiffness. Specimens of different mixing ratios of UV Electro and thinner were cast and subjected to tensile tests to quantify stiffness properties (see [8] for details). In a separate test, the stiffness of cast UV Electro without thinner was also determined. The results of these tests are discussed in the Results section below.

Printer Design

A modified desktop CNC milling machine (Zen Toolworks, Inc., www.zencnc.com, Concord, CA) was converted into a 3D printer by replacing the spindle with a custom extruder (Fig. 2). Computer models of the geometries to be printed were generated and converted to code that was sent to the 3D printer. For further information, including details of the design, description of the control software and parameters, and engineering drawings, see [8].

FIGURE 2.

Left: 3D printer with custom syringe extruder. Right: Illustration of syringe extruder. The housing firmly held the syringe in a vertical orientation while a lead screw attached to a linear actuator applied pressure to the end of the syringe plunger to extrude the silicone.

Print Preparation and Post-Processing

The silicone was prepared by thoroughly mixing a 10:1 ratio by weight of UV Electro base and its respective UV curing agent. Thinner (Silicone Thinner, Smooth-On, Inc.) was then added, with the amount based on the desired cured stiffness. Pigment (Silc Pig, Smooth-On, Inc.) was added to increase visibility of the silicone in the support matrix. This mixture was degassed and added to the syringe. The syringe was positioned into the custom extruder. A container with previously-prepared support matrix was placed on the bed of the 3D printer, and once the needle was positioned to the desired starting location within the support matrix, the print commenced.

Following the conclusion of the print, the container was carefully removed from the print bed and placed into a UV curing bed. The print was cured for ten minutes and then removed from the support matrix. Support matrix residue on the surface of the print was removed. The print was then placed into a bath of 70% isopropyl alcohol, lightly agitated for five minutes to remove any excess support matrix, and dried using a paper towel.

3D-Printed Cubes

To characterize the tensile properties of 3D printed silicone materials, six cubes (1 cm on a side) were printed using a 1:3 mixing ratio (UV Electro:thinner; i.e., one part by weight UV Electro to three parts thinner). Three control cubes were also cast using the same silicone mixture. Printing of the cubes took approximately 10.5 minutes. After post-processing, each cube was glued between two aluminum plates using silicone adhesive (Sil-Poxy, Smooth-On). The plates were mounted to an Instron 3342 Universal Testing System (Instron, Norwood, MA). The cubes were strained to 50% at a rate of 50 mm/min.

Each cube was mounted in one of two different orientations to test for anisotropy. These orientations are referred to as “parallel” and “perpendicular” based on the direction of the applied tensile force relative to the printed layer orientation direction. For the parallel (perpendicular) orientation, the force was applied parallel (perpendicular) to the printed layers. Three cubes were tested in each orientation. As the models were stretched, delamination of the printed layers was observed in both orientations; however, more separation occurred in the perpendicular orientation. Because of this layer separation, each model was only tested once and then discarded. Force-displacement data were used to generate stress-strain data. For further details about the tensile test measurement procedure, data processing, and analysis, see [8].

3D-Printed Vocal Fold Models

Vocal fold models were printed to determine whether their vibratory characteristics would be suitable for voice research. The outer profile of the printed models followed the geometry of the cover layer of the so-called “EPI” model described in [7] (see Fig. 3). The models were printed using a single silicone mixture, after which a simulated epithelial layer was added by subsequently pouring a thin layer of silicone over the models (i.e., as in [7]).

FIGURE 3.

Left: Parameterized VF geometry. Right: Profile used in the present research, obtained using the following dimensions: H = 8.4 mm, T = 0.1 mm, θ_1c = 50°, θ_2c = 5°, θ_3c = 90°, r_1c = 6.0 mm, r_2c = 0.987 mm.

As discussed below, the 3D-printed cube tensile test results revealed that the prints were anisotropic. Further, based on the printing paths, it was expected that the material would be transversely isotropic. In light of this, VF models were printed with two different layer orientations, horizontal and vertical, as illustrated in Fig. 4. It is evident in this figure that the printing orientation adjusted the plane of isotropy in each VF model, with the layers for the horizontally- and vertically-printed models lying within the sagittal and coronal anatomical planes, respectively. Additionally, separate models were printed using UV Electro:Thinner mixing ratios of 1:3 and 1:4.

FIGURE 4.

Top: Illustration of two VF models in their printing orientations. The outlined, red layer of each model indicates the first layer that was printed and the arrow indicates the build direction of subsequent layers. Bottom: The same models oriented for phonation testing, with the arrow indicating air flow direction.

After the models were printed, cured, and cleaned, the epithelial layer was added to the model. Similar to the process outlined by Murray and Thomson [7,19], this layer was applied by placing the model on a flat surface and pouring a thin layer of Dragon Skin (Smooth-on, Inc.), a stiff two-part silicone, over the model, allowing gravity to spread the Dragon Skin silicone into a thin layer, and then letting it cure. Excess Dragon Skin silicone around the base of the model was cut away. Only one epithelial layer was poured. The application of this layer had a slight surface smoothing effect and also added a degree of mechanical integrity to the models. As is customary with silicone VF models, talcum powder was applied to the models’ surfaces to reduce surface tackiness.

Using the setup shown in Fig. 5, the VF models were tested to acquire frequency and onset pressure data. Each model was mounted to an acrylic mounting plate using Sil-Poxy silicone adhesive. The mounting plate was secured to the testing table and placed against an acrylic plate to create a hemilarynx configuration. As discussed in the results section, in some cases a vertical restrainer was used to facilitate flow-induced vibration. Talcum powder was applied to the models to reduce surface adhesion between the model and the acrylic plate. Air flowed to the VF models via a rigid plenum and a 25.4 cm long subglottal tube. Vibration was induced by gradually increasing airflow until vibration began. Onset pressure was measured by a pressure transducer (6AF2G, Micro Switch, Morristown, NJ) mounted 5.1 cm upstream of the model. Model vibration was imaged using a high-speed camera (SC2+, Edgertronic, San Jose, CA) at 4000 fps and LED illumination (Model 900420H, Visual Instrumentation Corp., Lancaster, CA).

FIGURE 5.

Left: Side view of the hemilarynx configuration (not to scale). Right: Diagram of the red outlined area which shows the mounted VF, acrylic plate, and vertical restrainer.

RESULTS AND DISCUSSION

Suitability of UV-Cure Silicone for VF Models

Stress-strain data from specimens cast using different mixing ratios of UV Electro to silicone thinner are shown in Fig. 6 (top). The profiles are generally linear. Tensile moduli derived from these data are plotted in Fig. 6 (bottom) and listed in Table 1. Also listed in Table 1 is the modulus of UV Electro without thinner (measured in a separate test than the data shown in Fig. 6).

FIGURE 6.

Top: True stress vs. engineering strain curves for UV Electro mixed with different ratios of thinner as denoted in the legend. Markers denote every fifth data point. Bottom: Tensile modulus vs. mixing ratio found using a linear curve fit of the stress-strain data between 0 and 10% strain; numerical values are given in Table 1.

TABLE 1.

Tensile modulus values for UV Electro with different ratios of silicone thinner.

Mixing Ratio:	1:0	1:1	1:2	1:3	1:4	1:5	1:6	1:7
Tensile Modulus (kPa):	884.1	114.5	46.2	31.2	18.4	12.1	7.2	3.0

With the addition of thinner, the UV Electro modulus can be tuned by a factor of approximately 300 between thinner ratios of 1:0 and 1:7. These values fall within an acceptable range for voice research. For instance, Alipour and Vigmostad [20] found the average Young’s modulus of human VF tissue at low strains to be about 30 kPa in the longitudinal direction and 1 kPa in the transverse direction. Tensile modulus values of silicones used in VF models previously reported in the literature have been on the order of hundreds of Pa on the low end to more than 50 kPa on the high end (e.g., [7,15]). UV Electro can be formulated with higher mixing ratios for even more flexibility, but tensile tests at mixing ratios greater than 1:7 were not feasible because the exceeding flexibility of the specimens would cause them to sag under their own weight. Future rheometry tests could be conducted, however.

3D-Printed Cubes

Three 3D-printed cubes are shown in Fig. 7, each of which was printed with the same printing parameters and the same silicone and support matrix mixtures. Each resulted in an overall cube shape, although printing artifacts (such as the layer structures) and variations between prints are evident. Layer structure is inherent to extrusion-based 3D printing techniques. The variations between cubes are at least partially attributed to fluctuations in extrusion between the prints. It is anticipated that improved precision will follow further refinements to printer hardware, control software, and materials, as well as improved understanding of physical phenomena that are inherent in the process. Cubes with poor geometric accuracy (such as in Fig. 7b) were not used for tensile testing.

FIGURE 7.

Pictures of three cubes printed using the same printing parameters and silicone and support matrix mixtures. Cube A had the best dimensional accuracy. Cube B had poor dimension accuracy and was not used for tensile testing. Cube C had reasonable dimensional accuracy but slight layer variations.

Stress vs. strain data are shown in Fig. 8 for three types of cubes: cast, 3D printed and tested with the strain direction oriented parallel to the printed layers, and 3D printed and tested with strain perpendicular to the printed layers. At low strains (<10%), the curves appear to be quite linear. Tensile modulus values found by the slope of each curve are listed in Table 2.

FIGURE 8.

True stress vs. engineering strain data for cast cubes and cubes that were stretched parallel and perpendicular to the print orientation.

TABLE 2.

Tensile modulus values for the cast control cubes and the cubes printed in parallel and perpendicular orientations.

	Cast			Parallel			Perpendicular
Test #:	1	2	3	1	2	3	1	2	3
Tensile Modulus (kPa):	33.4	33.4	33.4	12.1	12.5	11.4	5.6	5.9	5.8

The results show that the 3D-printed cubes had fairly consistent stress-strain results within each type. The 3D-printed cubes were consistently less stiff than the cast cubes. The cubes that were tested in the parallel orientation had modulus values that were almost one-third those of the cast cubes, and the modulus values of the cubes tested in the perpendicular orientation were approximately half as large as those tested in the parallel orientation. The difference in moduli between the three cube types is attributed to a combination of uncured support matrix trapped within printed cubes and printing patterns used in fabricating the printed cubes. The results also show that layer orientation affected the stiffness, indicating that the 3D-printed cubes are anisotropic. These results are consistent with research that shows that 3D printing creates anisotropic structures [21–25].

The contrast between the two orientations demonstrates the anisotropy of the cubes. Because of the printing patterns, the cubes are likely transversely isotropic. Miri et al. [26] quantified the anisotropy of porcine VFs and found the stiffness ratio, defined as the ratio of the longitudinal to transverse modulus, to be between five and seven at low frequencies. In the present tests, the ratio of parallel to perpendicular stiffness was approximately two. One of the objectives of synthetic VF models is to imitate mechanical properties of a human VF as closely as possible; therefore, 3D-printed material that is also anisotropic has the potential to more closely mimic the mechanical properties of actual VF tissue than isotropic, cast silicone. Further tests would be needed to be performed to verify and more fully explore this possibility.

3D-Printed Vocal Fold Models

Figure 9 shows a single horizontally-printed VF model during printing. This figure shows how the support matrix supports the shape of the VF model throughout the approximately 47-minute print time. The final print shape satisfactorily matched the intended shape.

FIGURE 9.

Images of VF model being printed in support matrix.

Differences between models printed in different orientations (see Fig. 4) were observed. This is particularly evident in the inferior surfaces of the models shown in Fig. 10 in which it can be seen that the inferior and medial surfaces of the horizontally-printed model (with the printed layers in the sagittal plane) had less curvature, whereas the same surfaces of the vertically-printed model (layers in the coronal plane) were closer to the intended profile. This difference was attributed to the shape of the individual layers in the models. The layers of the horizontal model were rectangular whereas the layers of the vertical model were of the shape of the frontal profile shown in Fig. 3.

FIGURE 10.

VF models printed in different orientations. Dimensions are in mm. Shape differences of the models’ inferior and medial surfaces (as approximately outlined in red) are evident, with the vertical model better representing the intended profile. Blue dotted lines denote print direction.

Similar to the printed cubes, the printed VF models had small geometric errors. For example, the horizontal model in Fig. 10 had smaller medial-lateral thickness than the vertical model (8.3 mm vs. 8.6 mm). Further study is needed to determine the cause of this. Nevertheless, the model dimensions can be seen to be close to the desired values, and it is expected that process refinements will yield improved precision and geometric accuracy.

Another difference in print quality between the vertical and horizontal VF models was in a solid section of silicone that was observed to have formed during printing. The vertical models had a solid section of silicone that began at the bottom (where the print started) and extended part way up the model as shown in Fig. 11 (top). This also occurred in the horizontal models, but the solid section did not extend as far up the model (Fig. 11, bottom). Through experimentation, it was discovered that as a printed object grew in height, the layers near the bottom of the print began to connect and appeared to become solid silicone, nearly eliminating the layering pattern. Apparently because the horizontal print was shorter, the solid portion did not extend as high up in the print. Further research is needed to explore the mechanisms that drive this phenomenon.

FIGURE 11.

Top panel: Two views of the same vertically-printed VF model. Top left: View of the inferior surface (bounded by yellow lines) that shows the solid silicone region that extends approximately halfway up the model. Top right: Isometric view. Bottom panel: Images of a horizontally-printed VF model during printing. Layers at the bottom of the print are seen to connect to form a solid layer of silicone.

The results of the phonation tests are summarized in Table 3, and images of one oscillation period for each model are shown in Figure 12. The results demonstrate that life-sized, self-oscillating VF models can be 3D printed and that the models were durable enough to withstand oscillation. The frequency and amplitude of the models are comparable to those which human VFs experience during speech. The onset pressures were higher than the onset pressure for human speech (0.29 to 0.49 kPa, [28]) and those of previously-tested one-layer models (1.15 kPa, [29]). The elevated onset pressures of these models are partially attributed to the shape of the inferior surface, but mostly believed to be due to the internal structure and stiffness of the models being different than of actual VFs and other VF models. As this 3D printing process is refined and as the ability to print multiple models with stiffness variation is incorporated, it is anticipated that the geometry and material properties of the printed models will improve, which in turn will reduce the onset pressure.

TABLE 3.

Phonation properties for 3D-printed VF models fabricated using two silicone mixing ratios (1:3, 1:4) and different printing orientations (horizontal, vertical).

Mixing Ratio:	1:3				1:4
Orientation:	Horizontal		Vertical		Horizontal		Vertical
Model Name:	H11:3	H21:3	V11:3	V21:3	H11:4	H21:4	V11:4	V21:4
Onset Pressure (kPa)	4.0	4.2	3.8	3.9	3.2	3.9	3.2	2.6
Frequency (Hz)	222	228	181	214	230	218	129	148
Amplitude (mm)	0.84	0.76	1.26	0.53	0.75	1.43	2.20	1.00
VR Distance (mm)	8	8	-	-	6.5	6.5	8	8

FIGURE 12.

Frames from high-speed video for 1:3 models (top four rows) and 1:4 models (bottom four rows). Images show one complete period of oscillation. Differences in contrast between models are primarily attributed to differences in lighting and quantity of talcum powder on the model surfaces.

Each model was self-oscillated for approximately three minutes to gather phonation data, and there was no indication of model failure due to fatigue. Therefore, it is expected that the models could have oscillated longer. The horizontally-printed model H2_1:3 and vertically-printed model V2_1:3 experienced slight layer separation between the surface of the model and the epithelial layer while subglottal pressure was building and before vibration began. This separation was attributed to residual support matrix on the surface of the models which caused the epithelial layer to not adhere completely. This epithelial layer separation happened before the phonation data were recorded, so its effect on phonation before and after separation was not documented. However, after separation, the phonation properties seemed to be consistent over the course of multiple measurements of onset pressure and frequency during the approximately three-minute testing period. It is well-known that curing one silicone layer onto another already-cured silicone layer will result in the two layers bonding to each other if the already-cured silicone surface is clean. Surface contamination will inhibit bonding, and in this sense, the residual support matrix acted as a contaminant. Consequently, epithelial layer separation could likely be prevented by a more thorough cleaning of the model using isopropyl alcohol before applying the epithelial layer.

The vertically-printed 1:4 models had noticeably lower onset pressures and vibratory frequencies than the 1:3 models, whereas the horizontally-printed 1:4 models had slightly lower onset pressures than the 1:3 models with comparable vibratory frequencies. The difference in frequency between the vertical and horizontal models was more significant with the 1:4 mixing ratio than the 1:3 mixing ratio. The average frequency for the horizontally-printed 1:4 models was 62% larger than the average frequency for the vertically-printed 1:4 models, compared to a corresponding difference of only 14% for the 1:3 models.

The 1:3 and 1:4 horizontally-printed models and the 1:4 vertically-printed models would not self-oscillate without the addition of a vertical restrainer placed over the lateral portion of the models. The vertical restrainer was a thin piece of acrylic that was placed across the model at a specific distance and helped facilitate flow-induced vibration in a manner similar to that described by Zhang et al. [27]. This distance, VR, is illustrated in Fig. 5 and listed in Table 3. Zhang et al. used a vertical restrainer to change the mode of vibration of a single-layer VF model from being acoustically driven to aerodynamically driven [27]. A vertical restrainer could also reduce the pre-vibratory glottal gap, thereby facilitating vibration onset at lower pressures. Understanding the precise role of the vertical restrainers on the models’ vibrations described here, including the reasons why vertical restraint was needed in some cases, but not others, is an area for further investigation.

Some differences between the results of the horizontal and vertical prints were observed. The horizontally-printed models had a higher average vibratory frequency than the vertical models whereas the vertically-printed models had a larger average amplitude than the horizontal models. Additionally, the vertical models exhibited anterior-posterior vibratory asymmetry, particularly model V1_1:4 (see Fig. 12). This asymmetry was attributed to the silicone solidification at the lower regions of the prints (see Fig. 11) which resulted in anterior-posterior stiffness asymmetry. These observations suggest that future work is needed into the potentially significant effects of print orientation on the phonation properties of 3D-printed VF models.

CONCLUSION

A viable process for 3D printing self-oscillating synthetic VF models has been developed and demonstrated. The process is relatively low cost. The fabrication time for a newly-designed VF model is faster than current casting methods and no release agent is required. The results suggest that UV Electro may be a suitable replacement for Ecoflex 00–30, a material commonly used to fabricate cast synthetic VF models. Tensile test results on printed cubes revealed anisotropic mechanical properties as well as tensile moduli that were lower for printed vs. cast control cubes. The anisotropy of the prints is a potential advantage for voice research application, but further material testing is required to relate its mechanical similarities to human VF tissue. In addition, exploring the stiffness characteristics of VF models printed in different orientations would be of interest. Synthetic VF models were created that self-oscillated at frequencies and amplitudes similar to those of human VFs. Models were created using two different layer orientations, and through testing, it was shown that this layer orientation affected the properties of phonation. Onset pressures were higher than for human phonation and many other VF models, although it is expected that multi-layer prints and the use of softer materials will enable reduction in this parameter.

The technologies, processes, and application discussed herein are relatively new. As such, the primary purpose of this paper is to introduce this new approach for fabricating VF models. The intent is to show the potential for 3D printing self-oscillating VF models, hopefully stimulating further research into areas such as manipulation of layer orientation to achieve desired anisotropic characteristics, printing of multiple materials, and further material characterization. 3D printing opens new avenues for VF model design and fabrication, and it is anticipated that this process will be improved upon and used to create models with more anatomical realism and clinical relevance, with the ultimate goal of helping increase the understanding of voice production to improve voice care.